1. Field of the Invention
This invention relates to a digital/analogue conversion apparatus which converts a digital signal into an analog signal and also relates to applications using the same.
2. Description of the Related Art
U.S. Pat. Nos. 5,862,237 and 5,909,496 propose conventional examples of digital/analogue conversion apparatus which convert a digital signal into an analogue signal and applications which use the same where such a digital/analogue conversion apparatus coverts an audio signal into a plurality of digital signals and the audio signal is reproduced by using a plurality of speaker driver devices.
FIG. 1 of U.S. Pat. No. 5,862,237 shows that a digital serial audio signal is once converted into a plurality of digital signals using a serial-parallel converter and a decoder circuit. Here, the characteristic of this example is that the plurality of digital signals are converted so that they are weighted by the amplitude of the audio signal. In this way, a system is proposed which reproduces audio according to the amplitude of the audio signal when a plurality of speakers are driven by controlling the amount of current of an electric current supply of a plurality of driving devices in accordance with such weighting and driving the plurality of speakers.
FIG. 4 of U.S. Pat. No. 5,909,496 shows a digital serial audio signal is once converted into a plurality of digital signals using a serial-parallel converter and a decoder circuit as disclosed in U.S. Pat. No. 5,862,237. Here, the characteristic of this example is that the plurality of digital signals are converted so that they are weighted by the amplitude of the audio signal and the direction of the current of the drive circuits which drive the plurality of speakers is controlled using a specific single bit (MSB is a known example) among the plurality of digital signals. In this way, in addition to reproducing audio according to the amplitude of the audio signal when a plurality of speakers are driven by controlling the amount of current of an electric current supply of a plurality of driving devices in accordance with this weighting and driving the plurality of speakers, the drive circuits can be comprised by a simpler circuit.
In these conventional examples, because the serial-parallel converted signals are used as they are unadjusted as signals for driving a plurality of circuits the following problems occur: firstly, the manufacturing nonuniformity among current power sources of weighted drive circuits becomes a cause of non-linear noise, and second, when reproducing a digital signal the quantization noise which is generated during reproducing digital signals is superimposed as a noise component in an audible frequency band. Therefore, these examples suffer from difficulty in reproducing high definition audio signals.
In order to avoid the first problem a means is necessary for reducing the manufacturing nonuniformity among several drive devices.
In U.S. Pat. No. 5,872,532, a technology is proposed consisting of a selection circuit and an integrator for controlling the selection circuit as a means for reducing the nonuniformity between current supply sources which drive a plurality of speaker drive devices. According to this proposal, a signal which drives the plurality of speakers is input to a selection device and by controlling by a circuit which integrates once or more whether the plurality of speaker drive circuits have been used or not, the usage frequency of each of the plurality of speaker drive devices is integrated so that this integration result is maintained at a constant without depending on an input signal and the selection circuit is controlled. As a result, it becomes possible to reduce noise caused by the manufacturing nonuniformity among drive devices. Furthermore, the technology by which the nonuniformity among a plurality of drive devices is reduced is called a miss match shaping method.
FIG. 1 of U.S. Pat. No. 5,592,559 shows that an input serial audio signal is digitally modulated once using a sigma delta modulator and audio is reproduced by driving voice coils. While this conventional example proposes that a speaker in which two voice coils are driven in the positive and the negative directions using a three-valued signal which has been digitally modulated, technology for driving a plurality of two or more voice coils or reducing variation among a plurality of drive devices is not mentioned.
FIG. 3 of U.S. Pat. No. 7,058,463 demonstrates a proposal in which an input audio serial audio signal is digitally modulated once using a sigma delta modulator and over sampling and the output signal is pushed out to a higher frequency than an audible frequency band. Such technology in which a quantization noise is pushed out outside a certain frequency band in this way is called a noise shaping. In this conventional example, the quantization noise which occurs when reproducing a digital signal shifts to a frequency band higher than the audible frequency band using the noise-shaping method. Using this method the second problem in which the quantization noise is superimposed as a noise component of the audible frequency band is avoided.
Also, in order to avoid the first problem in which noise is generated due to the manufacturing nonuniformity among several drive devices, this conventional example proposes to introduce a miss-match shaping method which uses a selection circuit which is controlled by a DEM method (Dynamic Element Matching) using a pseudorandom signal.
However, a problem remains because even though a speaker drive circuit is driven without attenuating the quantization noise which is pushed out to a frequency higher than the audible frequency band by a digitally modulation using a sigma delta modulator and over sampling, the quantization noise which shifts to a higher frequency band is emitted from the speaker.
In addition, by simply switching the selection circuit by a DEM method using a random signal, white noise which is caused by such random signal is superimposed on to the audio signal which is reproduced. In order to avoid the problem of the noise caused by the manufacturing nonuniformity among several drive circuits, it is necessary to operate the switching of the selection circuit by the DEM method at a higher speed in addition to increasing the number of speaker drive circuits. The operation by the DEM method is given in detail in reference document “Delta—Sigma Converters” IEEE Press 1997 ISBN 0-7803-1045-4, section 8.3.3 and FIG. 8.5. The need for a high speed operation in a selection circuit is an serious weakness in the miss match shaping method which uses the DEM method. Furthermore, this weakness has already been pointed out in U.S. Pat. No. 5,872,532 and is widely known.
Pushing out the quantization noise which is generated by reproducing a digital signal to a frequency band above the audible frequency band by using a noise shaping method by digital modulation using a sigma delta modulator and over sampling is a technology which is generally well known. The relationship between the strength of noise which is shaped and the over sampling rate under a modulator order is shown in the formula in the reference document “Over sampling Delta-Sigma Data Converters” IEEE Press 1991 ISBN 0-87942-285-8, pp. 7 (22). Generally, in the noise shaping method, the effective strength of the quantization noise falls by 3 (2L+1) dB every time the over sampling rate is doubled where L is given as an order of a delta sigma converter. Therefore, in order to reduce quantization noise the over sampling rate must be increased or the order of the delta sigma modulator must be increased. On the other hand, when the over sampling rate is increased, it becomes necessary to operate the delta sigma modulator at a higher speed. In addition, when the order of the delta sigma modulator is increased the operation of the delta sigma modulator becomes unstable.
As stated above, in the noise shaping method in which a digital modulation is performed using a delta sigma modulation circuit and over sampling, the quantization noise which is generated by reproducing a digital signal is pushed out to a frequency band above the audible frequency band. Therefore, it is necessary to attenuate by a continuous time low pass filter (LPF) the unnecessary shaped quantization noise which is generated by the delta sigma modulation circuit or a component outside of the audible frequency band
FIG. 1(a) shows an example of a general system using a delta sigma modulation circuit. The unnecessary quantization noise or out-of-band component which is shaped and generated by a delta sigma modulator (100) is attenuated by a continuous LPF (101). Because over sampling is performed a low order LPF is enough. However, when a pass-band is narrow the decay time constant becomes larger and it becomes impossible to ignore the space occupied by the LPF mounted in a semiconductor integrated device.
There is a method for turning a delta sigma modulator into a multi-bit delta sigma modulator (110) as shown in FIG. 1(b) as a method for relaxing the required characteristics of the LPF which is placed after a modulator. In this case, by increasing the number of bits of the delta sigma modulator by one bit, it is possible to reduce quantization noise by 6 dB and therefore it becomes possible to relax the cutoff frequency characteristics of the LPF. However, by making a modulator into a multi-bit modulator, the circuit scale of an internal modulator increases significantly.
As another method for relaxing the required characteristics of an LPF, a method is proposed in which a Switched Capacitor Filter (121) shown in FIG. 1(c) is placed between the delta sigma modulator and the LPF. In this case, because a large capacitor is often required to reduce a cutoff frequency in addition to an OP amp in order to realize the Switched Capacitor Filter, this method suffers from a significant increase in chip area and consumption of power.
As another method for relaxing the required characteristics of an LPF, a method is proposed in which an analogue FIR filter (131) shown in FIG. 1(d) is placed between the delta sigma modulator and the LPF. In this method, the analog FIR filter is formed by analogically calculating and outputting each tap of the FIR filter. In this case, by increasing the number of taps, it is possible to increase the amount of attenuation of the out-of-band noise. This method which uses an analog FIR filter also has the effect of reducing deterioration in SNR by a clock jitter and is an effective method when using a clock signal with low accuracy or when using several clocks in the same chip.
Nevertheless, when a delta sigma modulator is turned into a multi-bit modulator, because only a delay element which forms an analog FIR filter and because the number of bits of the delta sigma modulator is required to be the number of cells of a segment type modulator multiplied by the number of taps, the circuit scale increases dramatically.